Merge branch 'main' into fix/return-star-respects-with-scope

Author: DeusData

graph-ui/src/App.test.tsx

@@ -0,0 +1,59 @@
+/* @vitest-environment jsdom */
+import "@testing-library/jest-dom/vitest";
+import { cleanup, render, screen, waitFor } from "@testing-library/react";
+import { afterEach, describe, expect, it, vi } from "vitest";
+import { App } from "./App";
+import { messages } from "./lib/i18n";
+
+vi.mock("./components/GraphTab", () => ({ GraphTab: () => null }));
+vi.mock("./components/StatsTab", () => ({ StatsTab: () => null }));
+vi.mock("./components/ControlTab", () => ({ ControlTab: () => null }));
+vi.mock("./lib/i18n", async (importOriginal) => {
+  const actual = await importOriginal<typeof import("./lib/i18n")>();
+  return { ...actual, useUiMessages: () => messages.en };
+});
+
+describe("App", () => {
+  afterEach(() => {
+    cleanup();
+    vi.unstubAllGlobals();
+    window.history.replaceState(null, "", "/");
+  });
+
+  it("shows the serving binary version", async () => {
+    vi.stubGlobal("fetch", vi.fn(async () =>
+      new Response(JSON.stringify({ lang: "en", version: "0.10.8" }), {
+        status: 200,
+        headers: { "Content-Type": "application/json" },
+      }),
+    ));
+
+    render(<App />);
+
+    expect(await screen.findByText("v0.10.8")).toBeVisible();
+  });
+
+  it("hides the version when the config has no string version", async () => {
+    const fetchMock = vi.fn(async () =>
+      new Response(JSON.stringify({ lang: "en", version: 108 }), { status: 200 }),
+    );
+    vi.stubGlobal("fetch", fetchMock);
+
+    render(<App />);
+
+    await waitFor(() => expect(fetchMock).toHaveBeenCalledWith("/api/ui-config"));
+    expect(screen.queryByTitle("Server version")).not.toBeInTheDocument();
+  });
+
+  it("hides the version when the config request fails", async () => {
+    const fetchMock = vi.fn(async () => {
+      throw new Error("offline");
+    });
+    vi.stubGlobal("fetch", fetchMock);
+
+    render(<App />);
+
+    await waitFor(() => expect(fetchMock).toHaveBeenCalledWith("/api/ui-config"));
+    expect(screen.queryByTitle("Server version")).not.toBeInTheDocument();
+  });
+});

graph-ui/src/App.tsx

@@ -33,8 +33,24 @@ function routeUrl(tab: TabId, project: string | null): string {
 export function App() {
   const t = useUiMessages();
   const [route, setRoute] = useState<RouteState>(readRoute);
+  const [version, setVersion] = useState<string | null>(null);
   const { tab: activeTab, project: selectedProject } = route;
 
+  useEffect(() => {
+    let cancelled = false;
+    void fetch("/api/ui-config")
+      .then((response) => (response.ok ? response.json() : null))
+      .then((config) => {
+        if (!cancelled && typeof config?.version === "string" && config.version) {
+          setVersion(config.version);
+        }
+      })
+      .catch(() => {});
+    return () => {
+      cancelled = true;
+    };
+  }, []);
+
   /* Normalize the URL on first load so it always carries the current route. */
   useEffect(() => {
     const initial = readRoute();
@@ -73,6 +89,14 @@ export function App() {
             <span className="text-[13px] font-semibold text-foreground/90 tracking-tight">
               Codebase Memory
             </span>
+            {version && (
+              <span
+                className="translate-y-px text-[10px] font-mono text-foreground/30"
+                title="Server version"
+              >
+                {version.startsWith("v") ? version : `v${version}`}
+              </span>
+            )}
           </div>
 
           {/* Tabs inline in header */}

internal/cbm/cbm.c

@@ -1,3 +1,7 @@
+/* Full declaration set for the same CBMArena, and it must precede cbm.h:
+ * internal/cbm/arena.h declares a subset and the two share the CBM_ARENA_H
+ * guard, so whichever is included first is the one this file sees. */
+#include "foundation/arena.h" // cbm_arena_init_sized
 #include "cbm.h"
 #include "arena.h" // CBMArena, cbm_arena_init/alloc/strdup/destroy
 #include "helpers.h"
@@ -817,9 +821,13 @@ static void cbm_error_regions_push(cbm_error_regions_t *acc, TSNode n) {
  * untouched.
  *
  * #1746: the Dockerfile grammar places that zero-width missing newline before
- * trailing horizontal whitespace rather than at raw EOF. Preserve the broad
- * exact-EOF rule above; only extend it past spaces/tabs when the missing token
- * is specifically a newline. */
+ * trailing whitespace rather than at raw EOF. Preserve the broad exact-EOF
+ * rule above; only extend it past blanks when the missing token is specifically
+ * a newline. */
+static bool cbm_is_blank_not_newline(char c) {
+    return c == ' ' || c == '\t' || c == '\v' || c == '\f' || c == '\r';
+}
+
 static bool cbm_is_eof_terminator_miss(TSNode n, const char *source, int source_len) {
     if (!ts_node_is_missing(n) || source_len < 0) {
         return false;
@@ -836,7 +844,7 @@ static bool cbm_is_eof_terminator_miss(TSNode n, const char *source, int source_
         return false;
     }
     for (uint32_t i = end; i < (uint32_t)source_len; i++) {
-        if (source[i] != ' ' && source[i] != '\t') {
+        if (!cbm_is_blank_not_newline(source[i])) {
             return false;
         }
     }
@@ -1186,11 +1194,27 @@ CBMFileResult *cbm_extract_file(const char *source, int source_len, CBMLanguage
     return r;
 }
 
-CBMFileResult *cbm_extract_file_ex(const char *source, int source_len, CBMLanguage language,
-                                   const char *project, const char *rel_path,
-                                   int64_t timeout_micros, const char **extra_defines,
-                                   const char **include_paths, const CBMMacroTable *macro_table,
-                                   const CBMReturnTypeTable *return_type_table) {
+/* Initial block for the per-file traversal scratch arena, chosen by measuring
+ * arena_grow on a 14k-file TypeScript tree: it fires on one file in 12,000 at
+ * both this size and at 1 MB, and on most files at 256 KB, where the two
+ * channel walks alone are exactly 262144 bytes. 512 KB therefore buys the same
+ * growth behaviour as 1 MB for half the resident block per worker. It is also
+ * exactly MI_LARGE_MAX_OBJ_SIZE in the vendored mimalloc
+ * (vendored/mimalloc/include/mimalloc/types.h:426, MI_LARGE_PAGE_SIZE/8 with
+ * MI_ENABLE_LARGE_PAGES defaulting to 1 at :115 and not overridden here), so
+ * the block is still bin-allocated from a large page. Growth is not free at
+ * this size for the same reason: arena_grow doubles to 1 MiB, which is above
+ * that bound and so a singleton OS allocation. One file in twelve thousand
+ * pays it, which is why the cost is accepted. */
+enum { CBM_EXTRACT_SCRATCH_BLOCK = CBM_SZ_512 * CBM_SZ_1K };
+
+static CBMFileResult *extract_file_ex_body(const char *source, int source_len, CBMLanguage language,
+                                           const char *project, const char *rel_path,
+                                           int64_t timeout_micros, const char **extra_defines,
+                                           const char **include_paths,
+                                           const CBMMacroTable *macro_table,
+                                           const CBMReturnTypeTable *return_type_table,
+                                           CBMArena *scratch) {
     // Allocate result on heap (arena inside for all string data)
     enum { SINGLE = 1 };
     CBMFileResult *result = (CBMFileResult *)calloc(SINGLE, sizeof(CBMFileResult));
@@ -1301,6 +1325,7 @@ CBMFileResult *cbm_extract_file_ex(const char *source, int source_len, CBMLangua
     // Build extraction context
     CBMExtractCtx ctx = {
         .arena = a,
+        .scratch = scratch,
         .result = result,
         .source = source,
         .source_len = source_len,
@@ -1429,6 +1454,7 @@ CBMFileResult *cbm_extract_file_ex(const char *source, int source_len, CBMLangua
                     // Build context for expanded source — extract only calls via unified extractor
                     CBMExtractCtx pp_ctx = {
                         .arena = a,
+                        .scratch = scratch,
                         .result = result,
                         .source = expanded,
                         .source_len = expanded_len,
@@ -1665,6 +1691,26 @@ CBMFileResult *cbm_extract_file_ex(const char *source, int source_len, CBMLangua
     return result;
 }
 
+/* Public entry. Owns the traversal scratch arena for the whole of one file's
+ * extraction: created here, handed to the body as ctx->scratch, destroyed on
+ * the way out. The body has seven early returns, so bracketing it in a wrapper
+ * is what keeps that to one create and one destroy. If the arena cannot be
+ * created, the body is handed NULL and the traversal stacks fall back to the
+ * result arena, which is what shipped before #1997. */
+CBMFileResult *cbm_extract_file_ex(const char *source, int source_len, CBMLanguage language,
+                                   const char *project, const char *rel_path,
+                                   int64_t timeout_micros, const char **extra_defines,
+                                   const char **include_paths, const CBMMacroTable *macro_table,
+                                   const CBMReturnTypeTable *return_type_table) {
+    CBMArena scratch;
+    cbm_arena_init_sized(&scratch, CBM_EXTRACT_SCRATCH_BLOCK);
+    CBMFileResult *result = extract_file_ex_body(
+        source, source_len, language, project, rel_path, timeout_micros, extra_defines,
+        include_paths, macro_table, return_type_table, scratch.nblocks > 0 ? &scratch : NULL);
+    cbm_arena_destroy(&scratch);
+    return result;
+}
+
 void cbm_free_result(CBMFileResult *result) {
     if (!result) {
         return;

internal/cbm/cbm.h

@@ -269,6 +269,10 @@ typedef struct {
                                         // pass_lsp_cross.c. Default false.
     bool requires_lsp_resolution;       // synthetic semantic candidate (for example an implicit
                                         // C++ operator). Never fall back to textual resolution.
+    bool callee_is_locally_bound;       // bare call foo() whose callee identifier is bound as a
+                                        // parameter of an enclosing function, so it cannot be the
+                                        // module-level foo. Python only today. Read by the
+                                        // weak-local-binding guard. Default false.
 } CBMCall;
 
 typedef struct {
@@ -589,6 +593,13 @@ typedef struct {
 
 typedef struct {
     CBMArena *arena;
+    /* Scratch for AST traversal, owned by the cbm_extract_file_ex call that
+     * built this context and destroyed when it returns. Nothing a
+     * CBMFileResult points at may be allocated here: `arena` is the result's
+     * own, and it outlives extraction by the whole pipeline (#1997). NULL in a
+     * context built without one, in which case the stacks fall back to
+     * `arena`. */
+    CBMArena *scratch;
     CBMFileResult *result;
     const char *source;
     int source_len;

internal/cbm/extract_calls.c

@@ -61,7 +61,8 @@ static const char *lookup_url_builder(const CBMExtractCtx *ctx, const char *name
 static int is_string_like(const char *kind) {
     return (strcmp(kind, "string") == 0 || strcmp(kind, "string_literal") == 0 ||
             strcmp(kind, "interpreted_string_literal") == 0 ||
-            strcmp(kind, "raw_string_literal") == 0 || strcmp(kind, "string_content") == 0);
+            strcmp(kind, "raw_string_literal") == 0 || strcmp(kind, "string_content") == 0 ||
+            strcmp(kind, "line_string_literal") == 0);
 }
 
 /* Strip surrounding quotes from a string, return arena-allocated copy */
@@ -2286,6 +2287,18 @@ static const char *extract_url_or_topic_arg(CBMExtractCtx *ctx, TSNode args) {
         if (strcmp(ts_node_type(arg), "argument") == 0 && ts_node_named_child_count(arg) > 0) {
             arg = ts_node_named_child(arg, 0);
         }
+        /* Swift wraps each argument in a value_argument that may lead with its
+         * label, so `data(from: url)` would otherwise yield the label `from`
+         * rather than the value. Step past a leading value_argument_label. */
+        if (strcmp(ts_node_type(arg), "value_argument") == 0 &&
+            ts_node_named_child_count(arg) > 0) {
+            TSNode val = ts_node_named_child(arg, 0);
+            if (strcmp(ts_node_type(val), "value_argument_label") == 0 &&
+                ts_node_named_child_count(arg) > 1) {
+                val = ts_node_named_child(arg, 1);
+            }
+            arg = val;
+        }
         const char *ak = ts_node_type(arg);
 
         if (strcmp(ak, "keyword_argument") == 0 || strcmp(ak, "pair") == 0) {
@@ -2973,6 +2986,41 @@ static bool python_receiver_is_exempt(CBMExtractCtx *ctx, TSNode receiver) {
     return false;
 }
 
+/* Name bound by one Python parameter node, or NULL when the shape binds none.
+ * Covers every binding form a `parameters` / `lambda_parameters` list produces:
+ * a bare `identifier`, the `name` field of default/typed parameters, and the
+ * identifier under a `*args` / `**kwargs` splat. A shape with no identifier (the
+ * bare `*` keyword separator) yields NULL and simply matches nothing. */
+/* True when the callee of a BARE Python call `foo()` is bound as a parameter of
+ * an enclosing function or lambda -- the bare-call counterpart of
+ * python_receiver_is_exempt above.
+ *
+ * A parameter binding shadows any module-level `foo` for the whole body, so
+ * resolving such a call to a project Function/Method by short name alone
+ * fabricates the edge BY CONSTRUCTION: `def _run_with_heavy_slot(run): run()`
+ * must not bind an unrelated `SatoriLive.run`. Unlike a receiver type this is
+ * decidable from the AST outright, with no flow analysis and no list of
+ * "generic-looking" callee names -- Python forbids `global` on a parameter, and
+ * a parameter is in scope for the entire body regardless of position.
+ *
+ * The answer is CARRIED BY THE WALK rather than recomputed here. The unified
+ * walk binds a def's or lambda's parameters when it opens that scope and
+ * unwinds them when it closes it, so this is an O(1) map lookup. Deciding it
+ * by ascending the tree instead -- with ts_node_parent() or with a copied walk
+ * cursor -- costs O(depth) per call, and since every level of f(f(f(...))) is
+ * itself a bare call, that is quadratic across the file: the 30,000-deep
+ * fixture in tests/test_stack_overflow.c turned it into a hang rather than a
+ * slowdown. An O(1) lookup needs no hop cap, so unlike a capped walk this can
+ * no longer fail open on deep-but-ordinary code.
+ *
+ * It still answers false if the walk's own tracking hit an allocation failure,
+ * which costs a suppression and never a true edge. */
+static bool python_callee_is_bound_parameter(CBMExtractCtx *ctx, WalkState *state,
+                                             TSNode callee_ident) {
+    const char *callee_name = cbm_node_text(ctx->arena, callee_ident, ctx->source);
+    return cbm_walk_python_param_is_bound(state, callee_name);
+}
+
 static bool is_objectscript_language(CBMLanguage language) {
     return language == CBM_LANG_OBJECTSCRIPT_UDL || language == CBM_LANG_OBJECTSCRIPT_ROUTINE;
 }
@@ -3045,6 +3093,19 @@ static TSNode objectscript_call_args(TSNode node) {
                                            : cbm_find_child_by_kind(macro_function, "method_args");
 }
 
+/* Swift models a call as a target expression plus a call_suffix, and its grammar
+ * declares no "arguments" field at all, so the generic field lookup finds
+ * nothing for every Swift call. Reach the argument list through the suffix
+ * instead. A trailing closure has a call_suffix with no value_arguments, which
+ * returns a null node and leaves the call without a string argument, as before. */
+static TSNode swift_call_args(TSNode node) {
+    TSNode suffix = cbm_find_child_by_kind(node, "call_suffix");
+    if (ts_node_is_null(suffix)) {
+        return (TSNode){0};
+    }
+    return cbm_find_child_by_kind(suffix, "value_arguments");
+}
+
 static bool node_has_token(TSNode node, const char *token) {
     uint32_t count = ts_node_child_count(node);
     for (uint32_t i = 0; i < count; i++) {
@@ -3597,11 +3658,18 @@ CBMInvocationDescriptor handle_calls(CBMExtractCtx *ctx, TSNode node, const CBML
             // (`accelerator.print()` must not bind MockAccelerator.print).
             // Imported receivers stay unflagged: module.function() is Python's
             // canonical cross-file call and the import map resolves it.
+            // A BARE Python call foo() whose callee is bound as a parameter of an
+            // enclosing scope cannot be the module-level foo, so short-name
+            // resolution would fabricate the edge (`def f(run): run()` must not
+            // bind SatoriLive.run). Distinct from is_method: there is no receiver
+            // here, so the weak-member guard cannot see this class at all.
             if (ctx->language == CBM_LANG_PYTHON && strcmp(ts_node_type(node), "call") == 0) {
                 TSNode fn = ts_node_child_by_field_name(node, TS_FIELD("function"));
                 if (!ts_node_is_null(fn) && strcmp(ts_node_type(fn), "attribute") == 0) {
                     TSNode obj = ts_node_child_by_field_name(fn, TS_FIELD("object"));
                     call.is_method = !python_receiver_is_exempt(ctx, obj);
+                } else if (!ts_node_is_null(fn) && strcmp(ts_node_type(fn), "identifier") == 0) {
+                    call.callee_is_locally_bound = python_callee_is_bound_parameter(ctx, state, fn);
                 }
             }
             // TS/JS/TSX receiver-aware guard (#592/#606 direction; same intent
@@ -3634,6 +3702,10 @@ CBMInvocationDescriptor handle_calls(CBMExtractCtx *ctx, TSNode node, const CBML
             if (ts_node_is_null(args) && is_objectscript_language(ctx->language)) {
                 args = objectscript_call_args(node);
             }
+            // Swift has no "arguments" field either; its args hang off call_suffix.
+            if (ts_node_is_null(args) && ctx->language == CBM_LANG_SWIFT) {
+                args = swift_call_args(node);
+            }
             if (!ts_node_is_null(args)) {
                 call.first_string_arg = extract_url_or_topic_arg(ctx, args);
                 /* #952: routes registered inside Laravel `prefix()->group()`